Assembly of a large earthquake from a complex fault system: Surface rupture kinematics of the 4 April 2010 El Mayor–Cucapah (Mexico) M_w 7.2 earthquake Open Access

Large surface ruptures that cascade slip along multiple faults offer an opportunity to learn about mechanical relationships between faults, along with factors that control the state of stress in the seismogenic crust. The 4 April 2010, moment magnitude (M_w) 7.2 El Mayor–Cucapah (EMC) earthquake activated slip on at least 7 different faults; many were unnamed and some were not even known to exist prior to the earthquake. Most of the faults have relatively short total strike lengths (<30 km), none of them control topography, and Holocene paleoscarps are rare. Individually each fault could have been thought to present a low seismic risk, and no one suspected that they had the physical links required for them to behave as a single fault system that was capable of producing a major earthquake. The fault system has strands that dip at very shallow angles (as low as 20°) and are likely to be severely misoriented relative to the inferred regional stress. Along its entire length, the complex fault system is located within 1–3 km of faults that can be considered the dominant strands of the plate margin in this area. These dominant faults include the Laguna Salada and Cerro Prieto faults, which have produced many major historical earthquakes and together accommodate to 90% of relative plate motion (Anderson and Bodin, 1987; Mueller and Rockwell, 1991; Bennett et al., 1996; Genrich et al., 1997). Therefore, the EMC rupture significantly tests our understanding of rock mechanics and failure criteria to explain why plate-margin shearing bypassed optimally oriented faults to become partitioned onto an adjacent complex fault network.

The EMC surface rupture is arguably the most complex ever documented on the Pacific–North American plate margin (Fig. 1). However, multifault ruptures that have complex kinematics and occur off of the main strands of the plate margin are not uncommon in the central portion of the plate margin, as is well demonstrated by the three most recent earthquakes in this region, including the M 7.2 EMC in 2010, M 7.1 Hector Mine in 1999, and M 7.3 Landers in 1992 (Fig. 2). We argue that the central domain of the plate margin is affected by several geodynamic processes, including regional redistribution of plate margin shearing, gravitational potential energy gradients emanating from the Great Basin, and the inherently three-dimensional strain of transtensional and transpressional shearing. We propose that these factors help explain the surprising and somewhat unpredictable nature of major earthquakes as well as the complex fault behavior observed throughout the central domain of the plate boundary.

In this paper we present field offset measurements, field and lidar (light detection and ranging) derived mapping, and kinematic analysis of the 2010 EMC surface rupture. We provide further insight into the deep complexity beneath the superficial simplicity of the rupture, which, when viewed from the far field (S. Wei et al., 2011), belies a complex array of faults, at places spread across a zone >3 km in width, that slipped together in this earthquake. North of the epicenter, within the Sierra Cucapah, excellent exposures of the surface rupture through bedrock and coarse alluvial sediments provide a wealth of slip data from which to infer how this fault system assembled to generate a major earthquake. We find that despite the great complexity of the surface rupture and its kinematic behavior, the overall strain released by this earthquake was consistent with interseismic deformation, which deviates slightly from expected plate motion, but is considerably oblique to other nearby historical ruptures such as the 1992 Landers earthquake. The overall consistency of strain, despite the variety of faulting involved, could indicate that fault rupture was guided overall by a relatively smooth pre-earthquake stress field.

The 2010 M_w 7.2 EMC earthquake is located in the central portion of the Pacific–North American plate margin, which contains the Big Bend segment of the San Andreas fault. The Big Bend is the prominent left-stepping restraining bend in the southern San Andreas fault (Hill and Dibblee, 1953) and is known to rupture in great earthquakes due to elevated normal stress (Sykes and Seeber, 1985). Faulds and Henry (2008) proposed that the Big Bend of the San Andreas formed due to the linking of plate margin shearing in the Gulf of California with the older portion of the San Andreas system established along the California coast. Although the timing of the onset of transtensional plate margin shearing in the Gulf of California is controversial, it can be no younger than 5.5 Ma (Oskin and Stock, 2003) and may be as old as 12.5 Ma (Fletcher et al., 2007). Regardless of the timing of its initiation, the southern San Andreas fault is not the only fault that has transected the mechanically competent Cretaceous batholith that exists as a continuous belt along the entire length of the plate margin. Rather, a complex network of faults as far south as the Agua Blanca fault operate together to accommodate strain compatibility between the two offset and kinematically distinct domains of plate margin shearing (Figs. 1 and 2). However, only ∼80% of the total relative plate motion is transferred to coastal California (Freymueller et al., 1999; Argus and Gordon, 2001) through this left-stepping network, and other fault systems in the central domain such as the eastern California shear zone and the Garlock fault transfer plate margin shearing throughout the Great Basin from the Walker Lane belt to the Wasatch front. In this study we define the central domain of plate margin shearing to include all faults that accommodate the redistribution of plate margin shearing from the narrow integrated transtensional fault network in the Gulf of California to the more regionally distributed and kinematically partitioned systems in the north. This central domain is here named the Big Bend domain and extends from the Garlock fault in the north to the Agua Blanca fault in the south (Fig. 2). We submit that the redistribution of plate margin shearing is only one of several geodynamic processes that combine to produce complex multifault ruptures like the EMC earthquake.

The EMC earthquake occurred in the southernmost portion of the Big Bend domain and represents the largest seismic moment release at this latitude of the Pacific–North American plate margin since 1892 (Fig. 2). The primary surface rupture and seismicity are distributed along a complex fault network that extends ∼120 km along an azimuth of 315° from the northern tip of the Gulf of California to the U.S.-Mexico border, which makes it the longest historical surface rupture for this domain. Triggered slip was observed on most faults in the immediate vicinity of the EMC rupture sequence, as well as on significant portions of active faults as far north as the Mecca Hills segment of the San Andreas fault, more than 160 km away from the epicenter (Rymer et al., 2011; M. Wei et al., 2011).

In the immediate vicinity of the 2010 rupture, two major fault systems have been recognized that control topography and are thought to accommodate most of the plate margin shearing at this latitude: the Imperial–Cerro Prieto fault to the east, and the Laguna Salada–Cañada David fault system (Fig. 3; e.g., Mueller and Rockwell, 1991; Bennett et al., 1996; Genrich et al., 1997; Axen and Fletcher, 1998) to the west. Both of these fault systems exhibit Holocene scarps and have generated at least three historical earthquakes that were similar in magnitude to the 2010 earthquake (Fig. 3; Anderson and Bodin, 1987; Mueller and Rockwell, 1991; Hough and Elliot, 2004). The Imperial and Cerro Prieto faults are located along the axis of the Mexicali basin and can be considered the dominant faults of the plate boundary at this latitude (Fig. 3). These two faults have linear traces, and are oriented with an azimuth of 315° ± 5°. Earthquake focal mechanisms and surface ruptures indicate that they accommodate strike slip and likely represent nascent transform faults (Faulds and Henry, 2008). Like other major transform faults in the Gulf of California, the Imperial and Cerro Prieto faults have an en echelon geometry and significant modern volcanic activity. The world’s second largest geothermal field is located at Cerro Prieto, where the two faults meet at an extensional stepover (Lira-Herrera, 2005). Modeling of global positioning system (GPS) geodetic data indicate 42 ± 1 mm/yr of slip on this fault system, which is ∼87% of the total between the Pacific and North America plates (Bennett et al., 1996; Genrich et al., 1997).

The Laguna Salada fault and Cañada David detachment fault, located to the west of the 2010 surface rupture, compose another important active fault system that controls subsidence of the Laguna Salada basin and uplift of the Sierra Cucapah and Sierra El Mayor (Fig. 3; Axen and Fletcher, 1998; Fletcher and Spelz, 2009). The Cañada David detachment is an active low-angle normal fault with west-directed (hanging wall moves west) tectonic transport. The Laguna Salada fault dips steeply to the west, accommodates oblique dextral-normal slip (Mueller and Rockwell, 1995), and essentially acts as an oblique ramp for the detachment (Fletcher and Spelz, 2009). Geologic and geodetic studies indicate that this two-fault system accommodates 2–4 mm/yr of displacement (Savage et al., 1994; Mueller and Rockwell, 1995). Therefore, together the Laguna Salada–Cañada David and Cerro Prieto–Imperial fault systems accommodate more than ∼90% of the full plate motion budget. The remaining slip is most likely to be accommodated by three other fault zones west of Laguna Salada: the Sierra Juarez normal fault, and the San Miguel–Vallecitos and Agua Blanca faults (Fig. 2), both of which are predominantly dextral strike-slip faults.

Perhaps the biggest surprise of the EMC earthquake is that it did not rupture any of the four well-known active fault systems at this latitude (Imperial, Cerro Prieto, Laguna Salada, Cañada David), but rather occurred on an obscure and, in part, unrecognized system of faults. Very few of these faults control topography, and slip in 2010 actually lowered much of the range crest of the Sierra Cucapah, which suggests that the fault system has had a much lower slip rate than the adjacent major faults. Likewise, faulting in the Colorado River delta occurred along a buried fault that failed to propagate to the surface as a discrete fault scarp due to the transfer of coseismic slip to more penetrative strain distributed throughout the thick sequence of water-saturated deltaic sediments (Oskin et al., 2012), as has been proposed for a similar rupture style in the New Madrid area (Missouri, United States; e.g., Guccione et al., 2000).

The epicenter of the 2010 EMC earthquake is located near the center of its rupture zone, and coincides with the boundary between two distinct kinematic faulting domains that have profoundly contrasting geomorphologic expressions (Fig. 3). To the south of the epicenter, the rupture cuts across the plain of the Colorado River delta and is herein named the Delta domain (Fig. 3). North of the epicenter, rupture follows an east-dipping stack of oblique-slip faults through the high portions of the Sierra Cucapah, which we refer to as the Sierra domain. Both domains accommodate dextral displacement. However, south of the epicenter the dominant vertical sense of offset is down to the southwest, whereas north of the epicenter the dominant sense of vertical slip is down to the northeast. Near the lateral limits of the rupture, fractures have a more northerly strike and show macroscopic right-stepping curvatures relative to the strike of the main rupture. The northern limit of the rupture is dominated by north-northeast–striking cross-faults, similar to those that rupture in the 1987 Superstition Hills earthquake sequence (Hudnut et al., 1989) but on a scale at least 10 times larger in area; this zone composes a third kinematic domain, which we call the Yuha Desert domain (Fig. 3). Here we describe the overarching characteristics of rupture in each domain, as well as relevant aspects of its geology. This is followed by a more detailed examination of the faulting in the Sierra Cucapah, where most field observations were obtained.

Vast fields of liquefaction and fracturing distributed across an area of ∼833 km² are the dominant expression of the EMC coseismic surface failure in the Delta domain (Fig. 3). The Colorado River delta has very low relief and an average elevation close to sea level. Most of this area is actively subsiding beneath a cover of late Holocene sediment that caps a thick sequence of late Neogene deltaic deposits. Sedimentary backfilling of fault-controlled basins like that in the Cerro Prieto stepover exceeds 5 km in thickness (Dorsey, 2010). Despite fast slip rates and the focusing of plate margin shearing on faults in this domain, the surface expression of faulting is not well defined (e.g., Biehler et al., 1964; Suarez et al., 1982). For example, the M 7.1 1934 Cerro Prieto and M 6.1 1980 Victoria earthquakes on the Cerro Prieto fault produced extensive liquefaction throughout the region, but geologists associated with later field surveys could not identify scarps with appreciable tectonic displacement and no continuous surface rupture was observed (Biehler et al., 1964; Suarez et al., 1982). Although it is clear that the EMC rupture did not occur on any of the main faults in the Delta domain (i.e., Imperial–Cerro Prieto faults), its surface expression is in many ways similar to that of previous events.

Despite the vast size of the zone of surface failure in the Delta domain, individual fractures have relatively short strike lengths, typically hundreds of meters, and they vary widely in orientation. Coseismic fracture arrays are commonly marked by extensive sand blows and subparallel anastomosing arrays of tectonic scarps are typically intersected by other sets of fractures (Fig. 4A). Many sand-blow fracture arrays have strongly curvilinear traces that show progressive changes in strike of more than 180°. Such arrays appear to be associated with shallowly buried sand bars deposited along meanders in ancient river systems (Fig. 4). Therefore, a significant proportion of fracture sets likely formed by dewatering of buried sediments due to vigorous coseismic shaking, and need not accommodate tectonic displacements or be located in close proximity to a buried fault with coseismic rupture. Curving zones of liquefaction and fracturing have also been associated with head scarps of lateral spreads, such as those associated with the New Madrid earthquake sequence (e.g., Penick, 1976). Nonetheless, some scarps have as much as 1 m of vertical offset and many arrays are arranged in an en echelon configuration. The sense of stepping in the en echelon arrays indicates a component of lateral shear that changes systematically with orientation. In general, northwest-striking en echelon arrays are more abundant and show a left-stepping geometry with dextral lateral offset, whereas north-northeast–striking en echelon arrays are right stepping and record sinistral lateral offset. These kinematics are consistent with north-south shortening and east-west extension, which is typical of the general kinematics of plate margin shearing in this area.

Two macroscopic fields of intense liquefaction have been identified from aerial surveys conducted within months of the EMC earthquake as well as Google Earth imagery generated over a period of two years after the earthquake (Figs. 3 and 4). In general, fractures and sand blows are much more common near perennial stream channels and the most active agricultural fields, which were likely to have been more heavily irrigated prior to the EMC rupture. It is interesting that the fields of liquefaction and fracturing do not cross the modern channel of the Colorado River. Although paleo–sand blows and paleo–mud volcanoes are observed in pre-2010 Google Earth satellite imagery, such features associated with the EMC earthquake are absent. Another feature that is absent from the main channel of the Colorado River in this area is perennial water flow, due its upstream diversion through a network of canals to other portions of the delta (Fig. 3). The region surrounding the main channel likely underwent the same amount of vigorous shaking as the adjacent areas with extensive liquefaction and fracturing. Therefore the lack of water-saturated sediments at sufficiently shallow levels in the subsurface is the most likely reason for the absence of ground failure, and the main channel, which used to be the wettest part of the delta during previous large earthquakes, is now the driest.

As was the case with the previous two large earthquakes in the southern Colorado River delta (Kovach et al., 1962; Biehler et al., 1964; Suarez et al., 1982), geologists working on the EMC earthquake through field surveys and conventional photographic imagery alone would not be able to recognize laterally continuous surface rupture controlled by discrete faults at depth. Nonetheless, a systematic sense of dextral west-down surface displacement (50–300 cm) was detected with synthetic aperture radar (SAR) interferometry and pixel offset tracking analysis of ALOS PALSAR (Advanced Land Observing Satellite, Phased Array L-band SAR; EORC—Earth Observation Research Center, JAXA—Japan Aerospace Exploration Agency) imagery along a continuous trace that extends ∼55 km to the southeast of the main shock epicenter (Fig. 3; S. Wei et al., 2011). Slip on the main faults probably does not reach the surface in the thick sediments of the Colorado delta. Waveform modeling and SAR pixel offsets suggest a relatively simple subsurface fault geometry in this region with coseismic rupture controlled by a single west-dipping master fault, the Indiviso fault (S. Wei et al., 2011). However, significant fault strands with an east-down component of coseismic slip and postseismic slip were also detected with ALOS PALSAR interferometry in the southern delta region (Nelson et al., 2013), and seismic reflection profiles across the southern delta demonstrate the existence of a major transtensional basin to the west of the Cerro Prieto fault, which is controlled by a family of west-dipping faults that sole into an east-dipping detachment (Chanes-Martinez, 2012). This kinematically and geometrically complex system of newly discovered faults underlies the vast region of coseismic surface failure in the southern Colorado River delta (Fig. 3).

Pre-event and post-event lidar surveys provided a way to directly calculate the coseismic changes in surface elevation and revealed displacements of as much as 50 cm in the vertical component across buried faults that could be traced laterally for 12 km (Oskin et al., 2012). In Oskin et al. (2012), the surface rupture in the delta was described as having an unusual style where kilometer-length, discontinuous curvilinear shear zones are associated with gentle, 100-m-wide ramps, across which the topography steps by ∼50 cm in elevation. The traces of these surface displacements coincide very well with the central segment of the ALOS PALSAR derived trace, but lidar provided even more detail into the geometric complexity and branching of faults at depth (Fig. 3). Fault splays typically have a left-stepping en echelon configuration and strike more northerly than the overall trace of the surface rupture, which suggests that they likely accommodate the dextral component of coseismic displacement (Fig. 3).

The fields of intense liquefaction do not symmetrically surround the lidar and SAR based traces of surface displacement, but instead are largely located in the downthrown block to the southwest (Fig. 3). This relationship is especially well displayed with the larger, southern field of liquefaction (Fig. 3). A seismic refraction profile collected recently as part of the Salton Seismic Imaging Project is located near the main shock epicenter and demonstrates the existence of a major fault in the subsurface that dips 50° to the southwest and projects toward the lidar and SAR based traces of surface displacement along the northeastern limit of intense surface fracturing and liquefaction (Ramirez-Ramos, 2013). Several studies have reported that stronger ground motions occur in the hanging wall than in the footwall of dipping faults (e.g., Abrahamson and Somerville, 1996; Oglesby et al., 1998), regardless of kinematics (Oglesby et al., 2000). The larger ground motions in the hanging-wall block arise from the trapping of seismic waves between the ground surface and the master fault (Oglesby et al., 1998, 2000; Ma and Beroza, 2008). Therefore, the distribution of ground failure in the Delta domain is not only controlled by the extent of fluid-saturated cohesionless sediments, but also by the southwest-dipping orientation of the buried master fault.

The northern field of liquefaction deviates significantly from the trace of the SAR based surface displacement and projects much farther to the north along the edge of the southern Sierra Cucapah, which suggests that it was at least partially controlled by rupture on a different fault (Fig. 3). A separate branch of surface displacement near the main-shock epicenter has yet been identified by remote sensing, but the north-trending salient in the field of intense liquefaction corresponds very well with a similar salient in the distribution of the EMC aftershock series (Hauksson et al., 2010). Based on wave-form modeling, S. Wei et al. (2011) proposed that EMC rupture initiated on a north-striking fault near the epicenter as an M_w 6.3 subevent with normal dip slip and later became transferred to the northwest-striking faults that control the main surface rupture in the Delta and Sierra domains. Armstrong et al. (2010) documented significant Quaternary uplift and proposed the existence of a major range-bounding normal fault along the eastern mountain front of Sierra El Mayor and southeast end of Sierra Cucapah. The northern segment of this east-dipping fault is likely to be the progenitor normal fault that initiated the EMC earthquake. Its surface trace most likely coincides with the western margin of the northern field of intense liquefaction, which projects toward late Pleistocene fault scarps along the southwestern side of Cerro Prieto (Fig. 3). If true, the distribution of liquefaction in the northern field also largely occurs in the hanging wall of a major fault system that accommodated coseismic rupture at depth.

In summary, several processes controlled the formation of fracture arrays in the Delta domain, including (1) tectonic slip on buried faults, (2) hydrofracturing associated with liquefaction of material in buried sand bars of the Colorado River delta, and (3) gravitational collapse along the margins of river channels, irrigation canals, and other linear topographic features. Fractures formed by the latter two processes are produced by severe shaking, and thus all three mechanisms are related ultimately to seismogenic slip on tectonic faults at depth. In this study we document the existence of at least three rupture-controlling faults in the Delta domain. The diffuse nature of surface fracturing and incomplete coverage of imagery taken immediately after the 2010 earthquake has inhibited a more systematic structural analysis of the rupture in this domain.

Most of the northern half of the main rupture is called the Sierra domain. Here the rupture propagated 55 km through an imbricate stack of east-dipping faults along the western slope of the Sierra Cucapah (Fig. 3). In contrast to the Delta domain, where only late Quaternary sediments occur at the surface, the Sierra domain exposes pre-Cenozoic crystalline rocks as well as Miocene and younger volcanic and sedimentary strata, forming a very high relief mountain range reaching more than 1000 m in elevation. Although much of the surface rupture occurred in remote mountainous regions, it is very well exposed with a plethora of offset geomorphic features. Therefore, it is in this domain where we focused the most comprehensive structural analysis and detailed mapping to characterize the geometry and kinematics of faulting, which is presented in detail in the following sections.

An overarching characteristic of the rupture in the Sierra domain is its structural complexity at the surface. The surface rupture has breaks along multiple fault strands, and kinematics are dominated by oblique dextral-normal slip with an overall east-down sense of shear, which lowered the high topography of the Sierra Cucapah (Fig. 5). As distance from the epicenter increases toward the north, rupture passes systematically to structurally lower strands in the imbricate fault stack. Structurally lower fault strands in the stack have systematically shallower dips and the structurally lowest Paso Superior fault is inclined at angles as low as 20°. Part of the complexity of the rupture can be attributed to interaction with oblique-slip detachment faults that caused the rupture to branch into numerous strands within a few kilometers of the surface.

In the Yuha Desert domain, surface rupture returns to the low-relief Colorado River delta, which is both structurally and geomorphologically distinct from the Sierra domain (Fig. 3). Faults in the Yuha Desert domain generally have a much more northerly orientation than those in the Sierra domain, and many strike northeast, nearly orthogonal to the strike of the primary rupture farther south. Surface rupture is commonly located along distinct paleoscarps that define the limit of uplifted and abandoned alluvial fan surfaces (Fig. 6; Rymer et al., 2011). The kinematics of slip vary systematically with fault orientation, with northwest-striking faults dominated by right-lateral slip, northeast-striking faults dominated by left-lateral slip, and north-south faults largely accommodating vertical slip (Rymer et al., 2011). Offsets on individual surface ruptures in the Yuha Desert domain are small, varying from 2 to 8 cm, with a northward-decreasing gradient (Rymer et al., 2011); slip is thus two orders of magnitude less than is observed in the Sierra domain. In subsequent discussion herein we document the fact that most faults in the Yuha desert domain are located in the upper plate of a detachment fault that controls the northernmost segment of primary surface rupture in the Sierra domain, and likely forms the structural floor of this domain (Fig. 6).

Rymer et al. (2011) interpreted that surface rupture in the Yuha Desert domain is best described as triggered slip and distinct from primary slip observed farther south. In addition to dramatically lower magnitudes of slip, the domain is characterized by a much greater abundance of antithetic northeast-striking sinistral faults, and fracturing is distributed over a much wider region (Fig. 6). Nonetheless, the horsetail-style pattern of faulting is consistent with regional stress acting at the end of a major surface rupture with a significant component of right-lateral slip (e.g., Kim et al., 2004). Aftershocks of the EMC earthquake have a very high density in this region, and the largest aftershock (the M_w 5.8 June 15, 2010 earthquake) is located in the Yuha Desert domain, which is also consistent with the loading of faults at the ends of a major rupture.

The presence of paleoruptures with meter-scale offsets suggests that this domain has accommodated cross-faulting near the lateral limits of other major prehistoric ruptures (Fig. 6). Based on new mapping, it is now known that the 1892 rupture on the Laguna Salada fault also has a northern limit in this domain (Rockwell et al., 2010). Directly south of the Yuha Desert domain, in the northern part of the Sierra Cucapah, there are several major topographic and structural features that have the same northeast trend as many of the fault strands in the Yuha Desert domain. The most prominent of these is the Cerro Centinela, which is a block of uplifted Mesozoic crystalline basement reaching more than 750 m in elevation with an elongated linear ridgeline oriented with an azimuth of ∼020° (Fig. 3). Other fault-controlled northeast-trending ridgelines with lower relief are observed southeast of Cerro Centinela (Fig. 3). We interpret these features to collectively represent the limits of previous surface ruptures through the Sierra Cucapah. Thus the 2010 earthquake was anomalous in that it propagated farther to the north, beyond these more prominent northeast-striking faults.

In the Sierra domain, EMC rupture activated slip on four different master faults that are separated by one branching intersection and two left-stepping accommodation zones (Fig. 7). From south to north, these are Laguna Salda fault, Pescadores fault, Puerta accommodation zone, Borrego fault, Paso Inferior accommodation zone, and Paso Superior fault (Fig. 7). In this section we summarize an extensive database of coseismic displacements as well as the orientation of master faults and scarps along the surface ruptures in the Sierra domain (Fig. 7). The master fault is the structure that has accommodated the finite geologic displacement, and its orientation in any given area is defined by the tectonic contact that juxtaposes distinct rock units from different fault blocks.

Master fault and scarp orientations were determined from outcrop measurements and three-point solutions of their traces on high-resolution airborne lidar-derived topography (Fig. 8). Three-point solutions were systematically collected along short, coplanar sections that transect topographic features such as channels and ridges. Individual scarps commonly occur in arrays of finite width that generally follow the master fault trace. However, the position of individual scarps varies along strike, and, in some cases, the entire coseismic scarp array is entirely located on one side of the master fault contact. Within any given array, individual scarps occur as distinct geometric and kinematic sets and thus some scarps were oriented oblique to both the scarp array and the master fault along which they propagated.

Rupture kinematics were characterized in the field by measuring the lateral and vertical components of offset markers across individual scarps (Fig. 9). The vertical component of slip was measured by projecting any displaced ground surface with a subhorizontal cutoff angle across the scarp (Figs. 10A–10C). Such markers were very common, which permitted us to make abundant reliable measurements of vertical offset. Offset markers that allowed a direct measurement of both the lateral and vertical components of slip included diverse geomorphic features such as bar and channel deposits, channel thalwegs, ridges, tread-riser intersections, shrub coppice dunes, as well as anthropogenic features that include tire tracks on a network of four-wheel drive roads, Highway 2, abandoned segments of Highway 2, boundaries of military rifle ranges marked by linear rows of painted boulders, animal tracks, and game trails (Figs. 9 and 10).

Total coseismic slip was determined by integrating individual measurements obtained in transects across the multiscarp rupture arrays. For each transect, total coseismic slip was calculated by a vector summation of the three components of slip (lateral, heave, and vertical) from each measurement (See  Appendix 1 for more complete description of conventions and assumptions used in these calculations.) Transects were designed to be as comprehensive as possible, but in some locations not all components of slip could be measured due to the heterogeneous distribution of slip markers and/or incomplete exposure of the scarps. Therefore, some transects only integrate the vertical component and other transects only integrate the lateral components. Scarps with <5 cm of offset were not systematically measured by all geologists who contributed to the data slip database. Therefore, our measurements of total coseismic slip represent minimum estimates.

There are 335 measurements that represent our best estimates of the lateral and/or vertical components of total coseismic slip. In order to visualize along-strike variations in kinematics, integrated coseismic slip measurements were plotted as lines that scale in length with slip magnitude, as shown in Figure 11. Envelopes were superimposed on the measurements to visually smooth out high-frequency variations in slip and connect measurements that are most representative of the total slip magnitude.

All of the main data sets presented in this paper are included as supplemental material, and we have plotted each of them as separate layers of a KMZ file that can be viewed with Google Earth (see Supplemental File 1¹). See Supplemental File 2² for geographic locations, view direction, and other information associated with all field photos used as figures in this paper. See Supplemental File 3³ for a compilation of orientation data for scarp-forming faults and master faults. See Supplemental File 4⁴ for field based measurements of coseismic slip across individual scarps as well as those integrated from transects across multiple scarps. See Supplemental File 5⁵ for definitions of attribute field names used in tables and KMZ of the above mentioned supplemental material.

The Laguna Salada segment is the southernmost rupture segment of the Sierra domain and is closest to the epicenter (Fig. 11). The Laguna Salada rupture segment extends ∼7 km from the edge of the Delta domain to the intersection with the Pescadores fault. At this branching fault intersection, most primary coseismic slip becomes partitioned onto the Pescadores fault and other faults of the east-dipping imbricate array (Fig. 11). However, minor rupture, not exceeding 30 cm, continues to the northern limit of the EMC rupture on the west-dipping Laguna Salada fault, and thus most of the length of the 1892 rupture (M_w 7.2) was reactivated in 2010 (Figs. 7 and 12).

Based on the location of the epicenter, the rupture is likely to have propagated to the northwest through the fault array in the Sierra Cucapah. However, slip on the southeasternmost Laguna Salada fault is divided into two splays that merge toward the northwest (Fig. 11). This merging of rupture in the direction of propagation may be due to the focusing of seismogenic slip from the more broadly distributed shearing in the Delta domain. The dip of the Laguna Salada fault is steep to subvertical along most of its length, and dip direction changes along strike (Mueller and Rockwell, 1991). The strike of the Laguna Salada fault within this rupture segment ranges from 285° to 310°, and thus this rupture segment has the most westerly striking faults in the entire rupture array.

Total slip reaches a maximum of 331 cm and is dominated by right-lateral displacement (Fig. 11). Some sections of the Laguna Salada fault have no measurable vertical offset. Where vertical offset is observed, it is generally <30 cm and changes polarity from northeast side down to southwest side down along strike. The most common offset features observed in this segment were recent bar and channel deposits located along the active drainage (Fig. 10D). Displacement estimates were difficult to determine in places where the fault array was subparallel to bar and channel morphology.

The Pescadores segment extends ∼15 km along strike and is the longest segment where rupture is controlled by and confined to a single master fault (Fig. 11). The trace of the Pescadores fault follows closely along the western flank of the topographic axis of the Sierra Cucapah, reaching the highest elevations and the most remote locations along the entire surface rupture. The Pescadores fault strikes 300° to 320°. In contrast to the Laguna Salada fault, the Pescadores fault shows no changes in polarity of dip and dips steeply to the northeast along its entire length (Figs. 5 and 11).

Although the strike of the Pescadores fault is very similar to that of the Laguna Salada fault, the kinematics of rupture undergoes a marked change at the intersection of the two faults and vertical slip increases and is systematically northeast-side down along the entire length of this segment (Fig. 11). Locally, there are southwest-down scarps, but these were always antithetic and minor in magnitude compared to the northeast-down slip. It is interesting that this sense of vertical slip is inverse to the modern topography and lowers the very highest part of the Sierra Cucapah along the entire length of the segment (Fig. 11). The Pescadores scarp array locally dams nearly every southwest-flowing drainage that crosses it, and shutter ridges formed by multiple events with displacements similar to the M 7.2 EMC earthquake are common (Fig. 5). In the Pescadores segment, right-lateral slip is 2–4 times greater than vertical displacements and net slip reaches 381 cm. Much of the Pescadores surface rupture is located along very steep slopes composed of coarse boulder deposits and colluvium (Fig. 5). Coseismic slope failure erased evidence for large sections of the scarp array and made finding offset markers difficult. Nonetheless, offset features such as channels, ridges, and alluvial deposits were common near large drainages, which is where most of the field measurements of slip are concentrated (Figs. 10G and 11). The Pescadores rupture segment is kinematically one of the most homogeneous of the scarp array.

One of the most surprising rupture-related phenomena found along this segment is an active fumarole that appeared in a wide alluviated valley near the southern end of the Pescadores fault (Figs. 11 and 13). The Pescadores fumarole constantly emits a prominent vapor cloud that can be seen from distances of several kilometers, and the vapor cloud occasionally rises above nearby peaks before evaporating into the dry desert air. In addition to vapor, the fumarole also extrudes a dusting of altered rock particles that blanket the ground around the vent (Fig. 13). The most active vent measures 1.5 m in diameter and steam was also emitted from fractures and through coarse clastic deposits within 5 m of the main vent. Penetrative ground cracking was observed around the vent, which is located within 10 m of the coseismic rupture (Fig. 13). The fumerole is located along a subtle right-stepping deflection of the rupture trace, and thus its location may be controlled by a releasing bend in the Pescadores fault. The fumarole site is ∼16 km from the Cerro Prieto geothermal field, and magnetotelluric profiles demonstrate the existence of shallow geothermal anomalies that approach the southeast margin of the Sierra Cucapah (Fig. 11; Cortes-Arroyo, 2011). Dilatation of rock volume associated with the rupture activity (e.g., Morton et al., 2012) as well as the subtle releasing bend geometry of the Pescadores fault likely provided conduits and mobilized fluids to produce the observed fumerole.

The Pescadores fumarole does not appear in long-distance photographs and was not otherwise observed in helicopter surveys performed in the days immediately following the EMC earthquake. Therefore hydrothermal activity at the site might not have initiated for days or weeks after the rupture. However, by 27 April 2010 the fumarole was vigorously active when field geologists (this study) first visited this section of the Pescadores fault. Colleagues at the Comisión Federal de Electricidad associated with the Cerro Prieto geothermal field were informed of the fumarole within months of its discovery and consider it to be a new geothermal field in Baja California (Alvarez-Rosales et al., 2012). Barnard (1968), who was the first to map and name most of the faults in the Sierra Cucapah, reported that the ground was hot to touch in some of the fault-controlled valleys in the southernmost part of the range. However, to our knowledge, historical activity of the fumarole has not been previously documented. Nonetheless, hydrothermally altered rock particles similar to those seen on the surface are found entrained in late Quaternary alluvial deposits around the fumarole, suggesting that it is a relatively long lived hydrothermal perturbation with previous periods of activity possibly related to other ruptures on the Pescadores fault.

The Puerta accommodation zone exhibits complex faulting where surface rupture is transferred in a left-stepping manner from the Pescadores fault to the Borrego fault (Fig. 14). These two master faults are only separated by an across-strike distance of 2–3 km, but the distance along strike separating points where slip is entirely consolidated on a single master fault is ∼10 km. Near the northern end of the Pescadores segment, the Pescadores fault bifurcates and rupture follows the structurally lowest strand for ∼3.5 km before abruptly terminating in a region slightly removed from the master fault trace (Fig. 11). Surface rupture with minor net coseismic slip appears on the Borrego fault within a distance of ∼3 km from the termination of Pescadores fault segment. However, slip on the Borrego fault within the northern half of the accommodation zone is discontinuous and kinematically dominated by dip slip, which reflects the partitioning of slip onto other structurally higher faults. Both the Borrego and Pescadores faults dip to the east and form the structural roof and floor of the accommodation zone. However, because the Borrego fault generally dips at a shallower angle than the Pescadores fault, they should intersect at depths of 10–12 km. Therefore, the two master faults are not spatially isolated, but rather they are likely hard linked at depth.

The Cascabel fault is the only other previously mapped structure to accommodate surface rupture besides the Borrego fault and Pescadores fault. The Cascabel fault is an antithetic fault in the east-directed system and bounds the northeast margin of a narrow rift basin in the hanging wall of the Borrego fault. Throughout the southern half of the narrow rift basin, the sequence of Pleistocene gravels and Palm Spring formation together with the graben-bounding faults have been tilted 20°–30° to the west such that this portion of the west-directed Cascabel fault has been rotated through vertical and now dips steeply to the east (Fig. 15A). Throughout the accommodation zone, the Cascabel fault records as much as 163 cm of oblique dextral and east-side-down slip. This sense of vertical slip reverses the west-down finite displacement of the fault, and thus is an excellent example of how faults change in kinematics as they rotate to new orientations during progressive shearing. Another example of mechanical interactions between the Casacabel and Borrego faults is the kinematic partitioning of coseismic slip. Although both accommodate east-down normal slip, the Borrego surface rupture in this accommodation zone has no lateral offset (Fig. 15B). Instead the lateral component of coseismic slip is partitioned onto other structurally higher fault strands like the Cascabel that that dip more steeply and sole into the Borrego fault.

Much of the evidence for the kinematics and distribution of surface rupture was erased by coseismic slope failure in a manner similar to that observed along the Pescadores fault. Coseismic mass wasting eroded many sections of the surface rupture and its deposits buried others (Fig. 16). The complex distribution of surface rupture among numerous discrete faults along with the lack of preservation has made it nearly impossible to characterize surface rupture in the accommodation zone by field observations alone. Instead, the distribution of surface rupture in the accommodation zone was largely documented by satellite image differencing (S. Wei et al., 2011). Many of the surface ruptures through the accommodation zone occur on faults that either do not juxtapose different rock units or do not significantly offset lithologic contacts.

A prominent gradient in surface displacement, as detected with COSI-Corr (co-registration of optically sensed images and correlation) processing of pre-event and post-event SPOT satellite images, extends across nearly the entire length of the Puerta accommodation zone (S. Wei et al., 2011). In other parts of the Sierra domain, such gradients typically correspond to ruptured master faults. Given the poor preservation of scarps in this segment, it is certainly possible that the gradient corresponds to a continuous coseismic rupture that extends ∼9 km. However, the gradient passes across, but does not appreciably displace a prominent intrusive contact marking the outer margin of the Cucapah granodiorite. Therefore, if this is a continuous fault, it likely has low finite offset (< 1 km).

The 2010 rupture on the Borrego fault extends >10 km, but because it is flanked by accommodation zones where slip is partitioned among multiple master faults, we define the Borrego segment to only include the central portion where slip is localized on a single master fault (Fig. 14). Despite being the shortest master fault segment, extending only ∼3 km along strike, the Borrego fault rupture has the most diverse range of geometry and kinematics. The strike of the Borrego fault varies from 320° to 015°, and its curvilinear trace likely reflects the linking of misaligned fault segments (e.g., Ferrill et al., 1999) and the development of megamullion corrugations (undulations of the fault surface) such as those that are commonly found on basin-bounding normal faults with large displacements (>5 km; e.g., Axen, 2007).

In this rupture segment, the Borrego fault zone is generally more than 50 m wide and dips moderately (as low as 40°; Figs. 14 and 17A). Due to the broad width and moderate inclination of the master fault, damaged and altered fault zone rocks are commonly observed along the entire mountain flank of the uplifted footwall block. Therefore, the Borrego fault segment can be classified as a mountain-flank fault, which distinguishes it from a mountain-front fault (e.g., Cichanski, 2000). Because the scarps from the EMC earthquake are distributed throughout the width of the fault zone and also dip at moderate angles, they form a distinctive pattern of surface rupture that we designate as a mountain-flank scarp array (Fig. 17).

The topography along the Borrego fault segment is not as extreme as observed in the Pescadores segment and Puerta accommodation zone and, in general, scarp traces are very well preserved. Nonetheless, alluvial fan surfaces are moderately inclined and the deposits are very coarse grained, and lateral displacements were best defined by offset modern channels. The coseismic kinematics of the Borrego rupture segment are dominated by oblique dextral northeast-down normal slip, which reaches 344 cm in magnitude (Fig. 14). Near the northern limit of the rupture segment, an abrupt change in strike to the northeast coincides with a reversal in the sense of relative lateral offset to sinistral and an abrupt increase in the magnitude of antithetic southwest down vertical offset (Fig. 14).

The Paso Inferior accommodation zone is located along the margin of the Laguna Salada lake bed; it extends ∼6 km along strike and has a width of ∼2 km (Fig. 18). In contrast to the Puerta accommodation zone, the complexities of faulting and the nature of slip transfer between discrete faults are well preserved and well exposed. The accommodation zone has very low relief and even the uplifted blocks of many fault strands are buried beneath modern alluvium, which suggests that this segment of the surface rupture is dominated by subsidence. The southern limit of the Paso Inferior accommodation zone is located at the intersection of a low-angle detachment fault with the central portion of the Borrego fault, which is where coseismic rupture bifurcates (Fig. 19). One splay follows the curvilinear trace of the detachment in a more western direction (Figs. 18 and 19). The detachment dips 20°–45° to the northeast and forms the westernmost strand and structural floor of the Paso Inferior accommodation zone (Figs. 18 and 19A). The other splay follows the northward continuation of the Borrego fault, which forms the structural roof of the Paso Inferior accommodation zone (Fig. 18). The northern limit of the Paso Inferior accommodation zone is located where coseismic rupture reconsolidates on a fault that we have named the Paso Superior detachment (Fig. 18).

The geometry of the Borrego fault changes markedly at its intersection with the southern detachment fault, and to the north it becomes much steeper (70°–84°; Fig. 18). The change in geometry coincides with a change in rupture array architecture, and the wide mountain flank scarp array observed to the south (Fig. 17) collapses to a much narrower array with a well-defined principal scarp (Fig. 20A). However, a significant component of antithetic southwest-down vertical slip is also observed immediately adjacent to the northern Borrego fault (Fig. 18), and more antithetic slip is spatially partitioned onto the Cascabel fault a few hundred meters to the east.

The kinematic and geometric diversity of faulting in the Paso Inferior accommodation zone surpasses that observed in any other segment of the surface rupture. Here, three main fault strands accommodate the transfer of slip from the Borrego fault to the Paso Superior detachment. The main fault strands show a range in strike of 300°–020° and dips that range from 20° to 84° (Fig. 18). Despite the partitioning of slip, the magnitude of coseismic slip actually increases to 388 cm on the northern Borrego fault (Fig. 18), and is the maximum observed on a single master fault. However, within ∼2 km of this maximum slip section, rupture on the Borrego fault dies out altogether and becomes transferred to structurally lower faults located to the west (Fig. 18). The detachment fault in the southernmost Paso Inferior accommodation zone is the structurally lowest fault strand and shows both right-lateral strike slip and predominantly east-side-down vertical slip (Fig. 18). A significant number of scarps with southwest-down antithetic slip were also observed in this array. Net coseismic slip on the detachment is very low (<10 cm) near the intersection, but increases to a maximum of 242 cm (Fig. 18). From this maximum slip location toward the north, coseismic slip progressively diminishes as it becomes transferred to a prominent west-side-down antithetic fault that forms a small graben in the region where they overlap (Figs. 18 and 20B). The trace of the west-dipping fault is located along the edge of uplifted hills of gravelly basin fill, and it also controls the western limit of a narrow horst block of Mesozoic crystalline rocks in the footwall of the Borrego fault (Figs. 18 and 20B). Surface rupture on this fault persists for more than 4 km, which is nearly the entire length of the accommodation zone, before it dies out and is transferred to a much shorter, west-side-down fault at the northern limit of the accommodation zone (Fig. 18). Both of these west-dipping faults are almost completely dominated by normal displacement that reaches a maximum of 194 cm (Fig. 18). Only minor amounts of right-lateral displacement were observed along short sections of this fault (Fig. 18). Ground cracks with as much as several centimeters of offset were observed between faults, but these were not systematically mapped throughout the accommodation zone. This subpenetrative fracturing represents the accumulation of off-fault strain, as suggested by modeling (Oskin et al., 2012).

In summary, many of the main fault strands in the Paso Inferior accommodation zone dip steeply to the west and are antithetic to the east-directed hanging-wall transport of the main faults in the Sierra Cucapah. The northern and southern extremes of the Paso Inferior accommodation zone are defined by east-rooting detachment faults, and we propose that despite the lack of continuous surface exposure, they are actually the same. Coseismic slip increases through the southern part of the accommodation zone, but then abruptly decreases as slip is transferred from the detachment to the west-dipping antithetic faults. In addition, the lateral component on many sections of west-dipping faults is relatively small. These two observations indicate that the coseismic slip did not completely propagate to the surface and that much of the missing slip may be the strike-slip component. Alternatively, the occurrence of penetrative ground cracking indicates that the missing slip may have been accommodated by penetrative off-fault deformation.

The EMC earthquake revealed the activity of the Paso Superior detachment fault, along which rupture extends 9.5 km. Previous mapping in this area had demonstrated the existence of isolated exposures of low-angle detachments that form the base of the sedimentary basin and they were all inferred to be young, but inactive (Axen and Fletcher, 1998; Chora-Salvador, 2003; Fletcher and Spelz, 2009). Not all of the known exposures of the basin-bounding detachment slipped in 2010. However, those that did rupture have a remarkably shallow inclination, as low as ∼20° (Fig. 18). All sections of the Paso Superior detachment accommodate strongly oblique dextral-normal slip. Therefore, this earthquake produced the first well-documented example of a seismically active low-angle normal fault that also accommodates significant lateral offset.

The Paso Superior detachment is the structurally lowest fault in the east-directed fault array of the Sierra domain. It juxtaposes Mesozoic crystalline basement with a thick section of boulder conglomerates and gravels that make up the sedimentary basin exposed over the full width (3–5 km) of the northern Sierra Cucapah in its hanging wall (Fig. 7). The fault zone of the Paso Superior detachment reaches thicknesses of as much as 100 m and is the widest of all master faults in the Sierra domain. In addition to penetrative gouge and breccia, chloritic alteration and coherent cataclasites are common in the fault zone; this indicates that they were unroofed from deeper levels in the seismogenic crust. All of these structural characteristics suggest that the Paso Superior detachment has the most net geologic offset of any fault in the east-directed array of the Sierra Cucapah.

Along most of the Paso Superior detachment, the vertical component of coseismic slip is slightly greater than the right-lateral component and, in general, right-lateral slip increases toward the north (Fig. 18). Total coseismic slip reaches 332 cm and is generally >250 cm along most of the segment. However, slip diminishes slightly near a small left step in the rupture array located south of Highway 2 (Fig. 18). Offset markers include channels, ridges, alluvial deposits, and a host of anthropogenic markers such as a network of four-wheel drive roads, Highway 2, abandoned segments of Highway 2, and boundaries of military rifle ranges marked by linear rows of painted boulders (Figs. 10E, 10F, 10H).

The southern Paso Superior detachment is not exposed, but the location of the blind structure is marked by the EMC rupture array, which follows the base of a series of hills containing uplifted gravel deposits along the margin of the Laguna Salada basin (Fig. 21A). To the north, the footwall crystalline block becomes well exposed where the detachment contact steepens to 45° in a ramp section that extends ∼0.3 km. Here, the structurally lowest scarps of the EMC array have the same inclination as the master fault and follow the trace of the detachment contact (Fig. 21B).

Like the Borrego fault, the Paso Superior detachment has a curvilinear trace that reflects the presence of megamullions. The EMC rupture array does not curve to conform to all of the changes in strike that define the megamullions, but instead splays into en echelon steps and shows distributed surface rupture over a width of nearly 1 km, which corresponds to the amplitude of the megamullions. Highway 2 is located near the axis of a synformal megamullion, and here the rupture array is entirely located in the hanging wall of the Paso Superior detachment, which is well exposed in a deep roadcut and dips <20° (Figs. 22A and 23). Scarps near the fault trace accommodate normal-sense dip slip, and nearly twice as much strike slip is partitioned across a 100–150-m-wide zone of cracking and minor faulting to the east (Fig. 22A).

North of Highway 2, the relief of the footwall block increases dramatically by 300 m adjacent to a second ramp in the detachment fault (Fig. 22B). This ramp is longer (∼1.5 km) and steeper (55°) than the one to the south, and much of the chloritic cataclastic zone has been cut out of the footwall. However, shallowly dipping foliated cataclasite crops out on the hanging-wall side of the EMC rupture and it is clearly refractured along discrete slip surfaces; this indicates that coseismic rupture likely propagated along the gently dipping fault zone before becoming diverted along the ramp near the surface.

Primary rupture in the northernmost Sierra domain leaves the Paso Superior detachment and expands into a zone of distributed surface fractures with minor offset through the Yuha Desert domain to the north (Figs. 3 and 6). Here the Paso Superior detachment curves strongly toward the west, and is not exposed north of its intersection with the Laguna Salada fault. Most of the faulting in the Yuha Desert domain is located in the hanging wall of the north-dipping Paso Superior detachment, which likely forms the structural floor of the Yuha Desert domain (Figs. 3 and 6).

In the Sierra Cucapah, the 2010 rupture propagated ∼55 km through a geometrically and kinematically complex array of faults, which we have subdivided into 4 master fault segments and 2 accommodation zones (Fig. 24). On the 4 master faults, total coseismic slip typically averages more than 2 m and individual measurements that reach 3–4 m are common. In the two accommodation zones, total coseismic slip measured on discrete faults diminishes; however, this is largely due to either lack of preservation of surface faulting or distributed penetrative off-fault deformation as documented in the Puerta and Paso Inferior accommodation zones, respectively.

The Borrego fault segment is located between two stepover accommodation zones. We have defined its limits to coincide with the location where coseismic slip diverges from a single master fault and becomes partitioned onto multiple master faults, the defining characteristic of the accommodation zones in this study. Therefore, portions of the lateral limits of the 2010 rupture on the Borrego fault are classified to form part of each accommodation zone. These sections are interesting because although slip is partitioned onto multiple faults, total slip magnitudes are much higher than other parts of the accommodation zone and more similar to total slip observed in other segments where rupture is controlled by a single master fault. Therefore, we identify two transition sections on either end of the Borrego fault, which has characteristics of both an accommodation zone and a master fault segment.

The greatest coseismic slip documented on a single fault is 388 cm, which was observed on a section of the Borrego fault in the southernmost transition region of the Paso Inferior accommodation zone (Fig. 18). However, because slip in this section is partitioned on 2 different master faults, the integrated coseismic slip is >4 m (Fig. 24). Rupture sections with the least amount of coseismic slip are hard to confidently identify with the results of field slip surveys. For example, our results indicate that total slip reaches a minimum of 0.48 m in the Paso Inferior accommodation zone. However, the existence of distributed ground shattering and off-fault strain suggests that total slip is much greater and possibly has much different kinematics than recorded by offset on discrete faults. Therefore, we do not have confidence in using field slip measurements to identify sections with the smallest amount of coseismic slip due to both penetrative off-fault strain and heterogeneous distribution of geomorphic markers, which may mask important components of the total surface displacement.

The ratio of lateral to vertical coseismic slip reflects the rake of the displacement vector. In general, lateral slip exceeds vertical slip in segments where rupture is controlled by a single master fault, but variations in the ratio of lateral to vertical slip are extreme. On the Borrego fault and Paso Superior detachment, the lateral to vertical slip ratios are slightly >1, whereas the ratio rises to 3 on the Pescadores fault and is almost 8 on the Laguna Salada fault. In contrast, slip on discrete faults in the Paso Inferior accommodation zone is dominated by vertical motions and has almost no lateral component.

In order to understand the extreme variations in the ratio of lateral to vertical slip of the EMC rupture in the Sierra Cucapah, we divided the array into the 49 separate kinematic sections, which we identified based on along-strike changes in geometry and kinematics (Figs. 11, 14, and 18). We determined the geometry of each section using Bingham statistics of observed fault orientations, and calculated the rake of slip based on the ratio of lateral to vertical slip averaged over the entire section. See Supplemental File 6⁶ for full documentation of the 49 geometric and kinematic rupture sections in the Sierra domain.

The strike of surface rupture weighted by section length ranges continuously from 290° to 010° with 31% of the data oriented 320° ± 5° (Fig. 25). The data are strongly skewed; most rupture sections strike with a more northerly orientation than the mode (Fig. 25). The average strike of all rupture sections weighted by length is 320°.

The projection of the vector sum of the lateral and dip-slip vectors onto a horizontal surface gives the slip direction. A rose diagram of slip directions weighted by section length demonstrates that 30% of all estimates have an azimuth of 310° ± 5° (Fig. 25). However, the slip direction data are bimodal and another smaller peak representing 5% of the data set has an azimuth of 80° ± 5°. Rupture sections with east-west–oriented slip directions are dominated by normal dip-slip kinematics. We hypothesize that in these sections, the partitioning of slip precluded the documentation of the full rupture kinematics. In general, the dominant slip direction (∼310°) is oriented ∼10° more westerly than the average strike (∼320°) of the rupture sections, which requires oblique dextral-normal slip (Fig. 25). We note that the slip direction of ∼310° is slightly oblique to the azimuth of relative plate motion, which is 320° (DeMets et al., 2010) at this latitude.

Using the fault-slip measurements and fault geometry, infinitesimal strain axes were calculated for each of the 49 rupture sections (Supplemental File 6 [see ^{footnote 6}]). The T-axis represents the principal extension direction of the infinitesimal strain ellipsoid, which is in general distinct from the least compressive principal stress (σ3) and the direction of relative plate motion. A rose plot demonstrates that T axes are relatively uniformly oriented and have an average orientation of 081° (Fig. 25). This orientation coincides well with the principal extension direction of infinitesimal strain determined to be 080° ± 2° from a trilateration geodetic survey across the Laguna Salada basin (Savage et al., 1994).

In summary, 2010 rupture in the Sierra Cucapah shows a diverse range of geometries and kinematics. However, the relative homogeneity of overall slip azimuths and T axis orientations indicates that regional stress is likely to be the dominant factor controlling kinematics.

The EMC earthquake demonstrated that the array of faults in the Sierra Cucapah have mechanical links that enabled the propagation of rupture through multiple faults. This study demonstrates that each fault has distinct geometry and kinematics, and that systematic along-strike variations in these parameters yield new insight into the mechanics of transtensional fault networks.

In general, the east-dipping fault array in the Sierra Cucapah can be thought of as an imbricate stack or fan of faults because individual faults dip progressively less steeply at lower structural levels, from subvertical for the Laguna Salada fault to as low as 20° for the Paso Superior detachment. Many of the east-dipping faults splay to the northwest from the subvertical Laguna Salada fault, which forms the southwest boundary to the imbricate fault system. The northernmost and structurally lowest rupture segment is controlled by the Paso Superior detachment, with a strike length of at least 9.5 km. Therefore, the EMC earthquake is the first modern rupture with dominantly strike-slip kinematics that is known to have occurred, in part, on a low-angle detachment fault.

Rupture through the Sierra Cucapah defines a path with an overall arcuate shape, and fault strikes vary systematically from 290° in the south to 340° in the north. However, the overall rupture is also characterized by a left-stepping en echelon geometry at scales that range from the macroscopic accommodation zones that span distances to 10 km (e.g., Puerta and Paso Inferior accommodation zones) to mesoscopic arrays with centimeter-scale fracture spacing. Therefore, individual fractures and rupture faults are oriented with a strike that is in general more northerly than the overall trend of the rupture, which can be approximated by a straight line with an azimuth of 312°. Although such an orientation is consistent with the simplified, geophysically defined rupture geometry (S. Wei et al., 2011), our data demonstrate that tectonic stress activated slip on planes with a wide spectrum of orientations and, in general, seems to have favored planes with a more northerly strike.

The en echelon steps in the path of the rupture not only allowed the rupture to have a more westerly strike than individual fractures and rupture segments, but they also formed the dominant mechanism for transferring slip to structurally lower master faults in the imbricate stack. In contrast to the ubiquitous left-stepping arrays, right stepping arrays are virtually absent and are only manifested in the macroscopic curvature observed at the lateral limits of the rupture in the Yuha Desert and Delta domains. The most likely mechanical explanation for the left-stepping arrays is that they accommodate dextral shear, which is the dominant kinematic component of the 2010 rupture. It is also likely that the left-stepping geometry and northward-curving fault tips reflect the obliquity between the optimum fracture orientation (i.e., one that maximizes the ratio between shear and normal stress) and the more westerly overall trend of the rupture.

Historic earthquakes in the Big Bend domain commonly involve multiple faults that link together in complex and often unanticipated ways (Fig. 2). For example, the three most recent large earthquakes in southern California and northern Baja California (1992 Landers, 1999 Hector Mine, 2010 EMC) all involved multiple faults with only parts of some faults rupturing along with others (Fig. 2). In Rockwell et al. (2000), it was demonstrated that the fault system behavior of the Landers earthquake is just one way that these individual faults can be linked and successive ruptures do not follow the same paths. The close proximity of the Landers (M_w 7.3) and Hector Mine (M_w 7.1) earthquakes surprised experts and provided new understanding for how large earthquakes cluster in space and time (Freed and Lin, 2001; Felzer et al., 2002; Oskin et al., 2008). Paleoseismic studies on faults that ruptured in the Landers earthquake confirmed the clustering hypothesis and indicated bursts of rupture events over short punctuated time periods followed by long quiescent intervals (5 k.y.) of strain buildup (Rockwell et al., 2000).

Nearly twice as many large historical earthquakes in the Big Bend domain are located on fault systems that are distinct from the dominant strands that have accommodated the largest finite slip (i.e., San Andreas, Imperial, and Cerro Prieto faults; Fig. 2). Large earthquakes in the Big Bend domain commonly nucleate in zones of diffuse strain such as the eastern California shear zone and the Sierra Cucapah terrane, which is sandwiched between fault systems that likely have an order of magnitude faster slip rates (Fig. 3). Many of the most active fault strands within the Big Bend domain, such as the San Miguel, Vallecitos and San Jacinto faults, are very young with low magnitudes of finite offset and have formed only within the past 1–2 m.y. (Fig. 2; Janecke et al., 2011; Dorsey et al., 2012; McGill et al., 2013). In contrast, the southern San Andreas fault has been ominously quiet for more than 300 yr (Fumal et al., 2002; Philibosian et al., 2011), and late Quaternary slip rates decrease by more than a factor of two south of its intersection with the San Jacinto fault (McGill et al., 2013). Although the southern San Andreas fault is considered overdue for a large earthquake, a lot of plate boundary slip has been taking place on the San Jacinto fault (Janecke et al., 2011) as well as on other faults with more favorable orientations (e.g., Nur et al., 1993; Du and Aydin, 1996). These issues underscore the challenge of accurately forecasting future earthquakes and there is general consensus that even in hindsight, many of the historic earthquakes in the Big Bend domain remain a surprise.

Longer term patterns of faulting in the Big Bend domain show even more complexity than is exhibited by historical ruptures and the large-scale kinematics of faulting show extreme spatial variations. Plate margin shearing from the Gulf of California bifurcates in the middle of the Big Bend domain and becomes partitioned into the San Andreas and Eastern California fault systems (Fig. 1; Faulds and Henry, 2008). In addition, the fundamental change in character of plate margin shearing from transtension in the Gulf of California to transpression on the San Andreas fault system also occurs in the middle of the Big Bend domain near the Salton Sea (Fig. 1).

Several key observations strongly indicate that both the redistribution of plate margin shearing and the reversal in the character of fault-normal strain reflects the influence of the ongoing collapse of the Great Basin. First, fault-perpendicular shortening is not restricted to the restraining bend segments of faults with more westerly strikes, but rather is observed along nearly the entire length (>1000 km) of the San Andreas fault system, and the orientation of subhorizontal maximum compressive stress changes systematically as fault traces become deflected around the southwest margin of the Great Basin (Townend and Zoback, 2004; Yang and Hauksson, 2013). The maximum horizontal compressive stress rotates more than fault strike and becomes more orthogonal to the San Andreas fault in central California north of the Big Bend domain (Fig. 1; Townend and Zoback, 2004). Second, along the southern margin of the Great Basin, Quaternary thrust faults and folds are distributed broadly throughout the Mojave block and accommodate regional north-south shortening and crustal thickening (Bartley et al., 1990). Geodynamic modeling has shown that, due to high mean elevation, heat flow, and anomalously buoyant mantle, the Great Basin has excess gravitational potential energy, which drives internal extension and causes the terrain to push outward on all of its margins (Jones et al., 1996; Flesch et al., 2000). Although some propose that extension in the Great Basin simply tracks the westward drift of the Pacific plate (Dixon et al., 1996; Wernicke and Snow, 1998), crustal shortening around its margins requires internal extension to be multidirectional and occurs at rates that exceed relative plate separation. Therefore, it is likely that the left-stepping deflection of the Big Bend has amplified through time due to bulging of the continental margin adjacent to the Great Basin. A third key observation is that along the southern limit of the Big Bend domain, faults with more westerly strikes such as the Agua Blanca fault accommodate transtensional shearing despite having the same geometry as restraining bend segments of the San Andreas farther north (Callihan, 2010; Springer, 2010; Wetmore et al., 2012). This demonstrates that the horizontal deviatoric compression around the margins of the Great Basin is not transmitted all the way across the Big Bend domain and that restraining bend kinematics are perhaps not as important as gravitational potential energy gradients for inverting the sense of fault-normal strain. However, both gravitational potential energy gradients and the rerouting of plate margin shearing require extreme spatial variations in the orientation and relative magnitudes of principal stress. These geodynamic processes operate in addition to relative plate motion, which helps to explain why faults that ruptured in the Landers earthquake have a much more northerly strike (to 40°; Hauksson, 1994) than those of the EMC event, even though both events are dominated by strike slip.

Perhaps the most important factor controlling the assembly of complex multifault ruptures in the Big Bend domain is the three-dimensional strain produced in both transtensional and transpressional shear zones. Shear zones that expand or contract perpendicular to their margins produce three-dimensional strain that is constrictional or flattening, respectively. In contrast, slip on a single fault accommodates only two-dimensional plane strain. Therefore multiple faults, typically more than four sets with distinct orientations and kinematics, are required to accommodate three-dimensional strain (Reches, 1983; Yin and Ranalli, 1992; Unruh and Lettis, 1998; Nieto-Samaniego and Alaniz-Alvarez, 1997). In the EMC surface rupture, we observe that slip was activated on faults with a wide spectrum of orientations and the rake of slip is directly controlled by fault strike. A geometrically and kinematically complex network of faults such as this is in fact required to produce the three-dimensional constrictional strain of transtensional shearing. In addition, because slip is commonly partitioned into multiple sets of faults that occur in the same rock volume, we hypothesize that different sets may be activated in different earthquakes. Therefore, the assembly of fault segments that link together to produce a large earthquake need not be replicated in consecutive ruptures.

The M_w 7.1 EMC of 4 April 2010 produced the most complex rupture ever documented on the Pacific–North American plate margin. Mapping and structural analysis of discrete surface rupture and distributed ground failure is consistent with rupture initiating on a north-striking fault and then propagating bilaterally to the northwest and southeast. Slip cascaded through a network of at least seven distinct master faults, which have a large range of diversity in geometry and kinematics. Fault segments that have end-member kinematics of either pure normal or dextral slip are relatively rare compared to those with oblique-slip kinematics and thus the overall slip direction is more westerly than the average trend (312°) of the rupture. Low-angle detachments control large segments of the surface rupture and played an important role in transferring slip through the fault network. Complex multifault ruptures that deviate from the from the direction of relative plate motion are common in the Big Bend domain due to the redistribution of plate margin shearing and the three-dimensional strain that is an inherent aspect of transtensional and transpressional shear zones.

This work was financed by CONACYT (Consejo Nacional de Ciencia y Tecnología) grant 81463, SCEC (Southern California Earthquake Center) grant 1697, and National Science Foundation grant EAR-0529922; the GEER (Geotechnical Extreme Events Reconnaissance) Foundation provided funding for initial field work. S. Leprince was supported in part by the Keck Institute for Space Studies and by the Gordon and Betty Moore Foundation. Part of this study was sponsored by the National Aeronautics and Space Administration (NASA) Earth Surface and Interior focus area and performed at the Jet Propulsion Laboratory, California Institute of Technology, under contract with NASA. Enlightening discussions with Paul Wetmore and Francisco Suarez helped refine ideas about tectonics of the Big Bend domain and the kinematics of faulting in the Colorado River delta, respectively. We thank John Galetzka, Kate Scharer, David Bowman, Roman Manjarrez, and Maria Oturno for help with field work. We also thank Jose Mojarro, Sergio Arregui, and Luis Gradilla for technical support.

Methods for integrating displacement from multiple measurements are described here. Rupture in the Sierra Cucapah occurred across an array of multistrand faults. Measurement transects oriented at a high angle to the array crossed as many as 19 individual scarps and fractures of varying orientations. Total coseismic slip was calculated by a vector summation of the three components of slip (lateral, heave, and vertical) measured in each scarp. The strike of every scarp was directly measured in the field and/or using their traces in map view, but in some cases scarp dip could not be measured because of a lack of three-dimensional exposure.

Therefore, in order to calculate the heave component of slip, a dip of 65° in the direction of vertical downdropping was assumed. The three components of slip in the scarp reference frame were converted to spatial coordinates: north, east, and down for the vector summation. The final vector summation was converted back to slip components relative to the local orientation of the master fault. Because geomorphic markers of vertical offset were the most abundant, this component of slip is the most accurate. In contrast, because it was not possible to directly measure the dip of every scarp in any given transect, the heave component of integrated slip is the least reliable. For this reason we do not report integrated heave calculated by vector summation. Instead, integrated heave is calculated using the integrated vertical displacement and the dip of the master fault, both of which are more robust measurements.

There are cases where the master fault dip could not be directly measured near a given transect due to a lack of exposure. In these cases we employed two conventions for approximating master fault dip. The first was to assume the master fault to have the same dip as that of the shallowest scarp-forming fracture measured near the transect. Field relationships demonstrate that scarp-forming faults are systematically inclined at higher angles than the master fault along which they propagated, and the disparity in inclination increases as the dip of the master fault decreases. Therefore, the shallowest dipping scarp is considered to be a maximum estimate of the true dip of the master fault. In rare cases it was not possible to systematically measure the inclination of individual scarps because of a lack of topographic relief to make three-point determinations. In these cases, the second convention was to assume the master fault dip based on structural context and known along-strike characteristics of the fault zone. Entries with assumed dips are clearly marked in the “Dip Source” field in Supplemental File 4 (see ^{footnote 4}).

Transects were designed to be as comprehensive as possible, but in some cases not all components of slip could be measured on all individual scarps due to the heterogeneous distribution of slip markers and/or incomplete exposure of the scarps. Therefore, some transects only integrate the vertical component and other transects only integrate the lateral components. Scarps with <5 cm of offset were not systematically measured by all geologists who contributed to the data slip database. Therefore, our measurements of total coseismic slip represent minimum estimates.